[Redirecting from the Ontology Summit list to Ontolog forum — followups should be directed solely to the latter list.]
On Dec 16, 2010, at 4:37 PM, Chris Partridge wrote:
Not every logician agrees with Chris (which I am sure is no surprise to anyone - including Chris).
Chris: That is assuredly true, but what I fear you have grossly misunderstood my previous post if you think that the citations you listed show some disagreement between me and the likes of Barwise and Moss. Nothing I said was in any way a criticism of non-wf set theory (as the quotes you included in your post would suggest). Indeed, as a postdoc, I had the pleasure of attending a seminar (along with Barwise) in which we went through a draft of what was to become Aczel's well-known text on the subject. (Larry Moss was in fact a fellow postdoc but, somewhat curiously in hindsight, I don't recall him attending the seminar.) I have been a big fan of the subject ever since. Barwise and Moss's terrific book Vicious Circles, in particular, contains both brilliant exposition and a series of impressive applications of the theory to very difficult topics in philosophy, semantics, and computation — topics whose complexity and subtlety require extraordinarily powerful mathematical tools. My post had nothing to do with such work; it had only to do with the adoption of non-wf set theory by the authors of ISO 15926.
So let me restate:
Point #1: While the full power non-wf set theories like AFA that are based on ZF (minus the axiom of foundation, of course) is necessary for the sorts of complex applications taken up in Vicious Circles and the like, such set theories are (as Barry had already pointed out) massive overkill for ISO 15926. (In particular, I assume that the ontology has no need for the entire class of transfinite ordinals, power sets of uncountable sets, and so on.)
Point #2: On the assumption that it is a "theorem" of ISO 15926 that (a) THING is a class and (b) everything is a member of THING, ISO 15926 is flatly inconsistent with ZF-based non-wf set theories like AFA, in which it is a theorem that there is no such class as THING.
Unless the authors of ISO 15926 don't care that their ontology is provably inconsistent, they need a very different class theory.
Chris Menzel
For example,
http://www.amazon.co.uk/gp/product/1575860082/ref=wms_ohs_productVicious Circles: On the Mathematics of Non-Wellfounded Phenomena - Jon
Barwise (Author), Lawrence Moss (Author)
Product Description
Circular analyses of philosophical, linguistic, or computational phenomena
have been attacked on the assumption that they conflict with mathematical
rigour. Barwise and Moss have undertaken to prove this assumption false.
This volume is concerned with extending the modelling capabilities of set
theory to provide a uniform treatment of circular phenomena.
Jon Barwise (http://en.wikipedia.org/wiki/Jon_Barwise ) argues that the
circularity found in non-well-foundedness is common in the real world - i.e.
it is a common requirement. As such it would make sense to include it on
your foundation.
While personally not finding all his examples persuasive, I think the
general point that it is a requirement is well made.
Another interesting recent (technical) book on a similar the same topic is http://www.amazon.co.uk/gp/product/0199276439/ref=wms_ohs_product - Absolute Generality - Agustín Rayo (Editor), Gabriel Uzquiano (Editor). I believe the introduction can be found somewhere on the net.
Chris
-----Original Message-----
On Dec 16, 2010, at 2:24 PM, Matthew West wrote:
-1 for 15926, with arguments:
http://ontology.buffalo.edu/bfo/west.pdf
Which are answered in: [THIS LINK]
Barry's criticisms of the use of a non-well-founded set theory like Aczel's
AFA are on the money. He notes that it is a greatly overpowered for the
needs of the document; it entails, among other things, the entire massively
infinite hierarchy of transfinite numbers. (How massive? So massive that
there is no transfinite number big enough to number them.) Moreover,
ironically, AFA and its like are in a sense underpowered as well for the given
task. Notably, as I understand the document, THING is itself a class that
contains, well, everything. The existence of such a class (understood as a
non-wf set) is flatly inconsistent with non-wf ZF spinoffs like AFA.
Bottom line (as John Sowa likes to say): The underlying class theory of
the document needs to be thrown out and rethought completely.
Chris Menzel